Even though it seems crazy that a game is controlled in this way it might be for 3 reasons:
1) Blizzard has a patent registered for this. (We know it and you have it above in my previous post)
2) If you think about it is not crazy because: You bring together those who spend little or those who have recently reduced their expenses with statistically unfavourable matches. This is because you encourage him to spend to upgrade or switch deck. But when someone has bought, you make him face favourable fights to make him complacent and increase satisfaction. Therefore, he will remember the satisfaction of the expense and he will be more inclined to pay next time.
3) We know for sure that Blizzard uses this kind of algorithm in mercenaries, because they said it. In mercenaries these algorithms are based on level, rank, equipment and so on in order to match the opponent. If you remove equipment and you add the average money spent in time they also have the algorithm for constructed.
The technology behind it is ultimately very simple (it is only math indeed), it can be done and it is done in other games.
In conclusion, a company like Blizzard that is very money-centric is not absurd that it is using a tool like this. I would not be surprised.
If the purpose of this algorithm is to force players to spend more money, then why has it been perfectly possible to hit legend with only face hunter since Descent of Dragons came out?
Well, this is because they had to balance between free-to-play and people who spend. Players are both. Anyway, it is not impossible to think that a face hunter player will find more matches against taunt druid, shadow priest or control warrior (that are not favourable matches) than a ramp druid for example.
These assumptions are pretty easy to check anyway. We need one person that plays around 100 matches or more with the same deck. Every deck is fine for this experiment.
If the numbers of his fights are aligned with the numbers in the attachment, everything is fine. If his numbers are not aligned with the numbers in the attachment we can assume that HS is controlled.
Well, this is because they had to balance between free-to-play and people who spend. Players are both. Anyway, it is not impossible to think that a face hunter player will find more matches against taunt druid, shadow priest or control warrior (that are not favourable matches) than a ramp druid for example.
These assumptions are pretty easy to check anyway. We need one person that plays around 100 matches or more with the same deck. Every deck is fine for this experiment.
If the numbers of his fights are aligned with the numbers in the attachment, everything is fine. If his numbers are not aligned with the numbers in the attachment we can assume that HS is controlled.
With an experiment like this, there is proof.
You need two people at the same rank (and possibly same server - not entirely sure how much variance there is based on that) with roughly the same MMR, realistically - the large datasets show variance between ranks. In practice, this means two legend rank players with a similar rating and different decks.
Also, even just looking at purely the classes, I'd look at a bigger sample than 100 - more like 2-300. I presume you're not planning on looking at 40-50 deck types with 100 sample size, for what should be obvious reasons!
No. You need one person and the MMR is not relevant. You are testing the meta. Therefore if in 100 matches he will not find (with always the same deck) 23,6% of druid (or around this number) something is strange, otherwise, everything is fine.
No. You need one person and the MMR is not relevant. You are testing the meta. Therefore if in 100 matches he will not find (with always the same deck) 23,6% of druid (or around this number) something is strange, otherwise, everything is fine.
But the meta will shift between different MMRs. Always has done. Top 100 legend meta will be significantly different to that in, say, silver 10 with a MMR that reflects that. Heck, it'll be different between bottom of legend more fun decks and D5-1 - in this case because of a difference in philosophy for the players.
People say that they play aggro until a certain point for faster and more consistent grinding, as an example.
The core issue is that you're matched based on MMR, which is invisible pre-legend - and that MMR influences your metagame, thus needs to be controlled for. 80%+ of actual science, in my experience, is controlling variables, cutting datasets, and similar.
You don't have to leave meta to shift. You have to try it during the same meta. If you let the meta shift during the experiment the whole experiment is pointless. Even though the experiment seems long, it is quite fast. Take a random deck (face hunter for example) and match the opponent and concede 100 times. Are your opponents aligned with 23,6% druids?
Yes - So HS is not controlled
No - HS is controlled
In Vicious syndicate, there are data about different ranks. So you have to choose one accordingly. Anyway, there are some differences, if you check the data, but nothing astonishing. 23,6% druid is the average in all ranks, so it is a reasonable benchmark.
That theory is wrong as others have mentioned (it’s just you being on a losing streak). The way to confirm is: don’t switch decks. Run the same deck for 50-100 games and keep track of your matchups (use a deck tracker). It will prove to you that what you think is happening, is not actually happening.
HSReplay tracks every single game played by it's users. Well over a thousand games minute.
At no stage have they been able to track any kind of pattern in the match-ups and if there were any it would be quite obvious given how many games they record.
People continually post of this forum about being matched up badly when changing decks, seeing fixed RNG working against them, seeing card draws being either too lucky or unlucky to be believable...
But it's all based on annecdotes at worst, or a short run of tracked games in their personal log at best - neither of which amounts to dick all.
Of course activision Blizzard have IP on all sorts of algorithm, they are heavily in the loot box box business, across nearly all their games. That doesn't mean some of the more nefarious ones are in use in hearthstone - and the second they were people would see.... I mean, err, people that are not so susceptable to conspiracy and paranoia would see...the rest of you see it already when it isn't even there - there isn't much helping you, you've been told so many times now.
You don't have to leave meta to shift. You have to try it during the same meta. If you let the meta shift during the experiment the whole experiment is pointless. Even though the experiment seems long, it is quite fast. Take a random deck (face hunter for example) and match the opponent and concede 100 times. Are your opponents aligned with 23,6% druids?
Yes - So HS is not controlled
No - HS is controlled
In Vicious syndicate, there are data about different ranks. So you have to choose one accordingly. Anyway, there are some differences, if you check the data, but nothing astonishing. 23,6% druid is the average in all ranks, so it is a reasonable benchmark.
There are multiple difficulties with that approach.
Yes, you can't let the meta shift - but you are looking at 'all ranks' in the VS data. Click into the rank breakdown and you will see that the overall 23.6% is all over the place: from 15% at GSB to 38% at 1k legend. There is a 4% difference between the Diamond brackets. So picking a random rank will not reproduce the same value.
You also cannot try to repeat last weeks results this week because the meta is shifting between weeks (which you can see in the different reports).
The meta is probably also shifting on a daily basis (weekdays - weekend), time of the day (ladder grinders during the day or casuals in the evening) and MMR (the skill to pilot certain decks).
So ideally you would want to break down the VS data to that level to have an accurate goal or in absence of that granularity, play enough games to cover all levels with a large enough sample size.
And then you have to consider the observer effect of you injecting games: if you beat players they might change their deck, if you concede they gain rank and move out of your bracket scope with possible knock-on effects. On top you are gaining or losing MMR, which can put you in a different meta.
So yeah, you want two players with different decks that queue at the same time into the same local meta to generate two data sets that you can then compare.
Yes. 4% difference is nothing. I don't know why you guys like complicated things. Mine is a pretty direct approach. No doubt that it is correct. Sure it is not perfect. But it is easy, correct and FAST (your models are not fast).
Sure if you make the model more complicated the result is better it is obvious but pointless. You can make it with 10 players and the model is even better and then?
Pretty easy and fast. Same deck. Concede 100 times. Are your data aligned with that of Vicious?
Yes - game is not controlled
No - game is controlled
Blizz has a patent for these things, so this is an easy way to check if they are using it in HS constructed.
I don't know your experience in data analysis, I have it for big investment banks and I can tell you that 100 is a good sample size. Most of the decisions are generally taken with a lower sample size (because indeed most of the time there is no possibility to have a 100 sample size). So, nothing strange for me.
you are so angry at everything that all your reasoning is biased and grounded on a prejudice ("Blizz... i hate them, they are the EVIL!!!"). you start with the conclusion and then try to find proofs to support your theory. scientific approach is the other way around mate. this, above, is called: paranoia.
You can test it yourself relatively easily. Build a deck of absolutely bad cards, don't use any strategy. Just build a deck yourself. Your next opponent will also play an off-meta deck. Almost always works. It's just an extra screw to bring some "balance" into the game.
you are so angry at everything that all your reasoning is biased and grounded on a prejudice ("Blizz... i hate them, they are the EVIL!!!"). you start with the conclusion and then try to find proofs to support your theory. scientific approach is the other way around mate. this, above, is called: paranoia.
I am chill. We are talking about an approach to test it. You are out of topic. Please read better next time.
You can test it yourself relatively easily. Build a deck of absolutely bad cards, don't use any strategy. Just build a deck yourself. Your next opponent will also play an off-meta deck. Almost always works. It's just an extra screw to bring some "balance" into the game.
Or it could just be that people who are at a certain MMR... play off meta decks.
Leaving aside all the jokes on the topic.
Even though it seems crazy that a game is controlled in this way it might be for 3 reasons:
1) Blizzard has a patent registered for this. (We know it and you have it above in my previous post)
2) If you think about it is not crazy because: You bring together those who spend little or those who have recently reduced their expenses with statistically unfavourable matches. This is because you encourage him to spend to upgrade or switch deck. But when someone has bought, you make him face favourable fights to make him complacent and increase satisfaction. Therefore, he will remember the satisfaction of the expense and he will be more inclined to pay next time.
3) We know for sure that Blizzard uses this kind of algorithm in mercenaries, because they said it. In mercenaries these algorithms are based on level, rank, equipment and so on in order to match the opponent. If you remove equipment and you add the average money spent in time they also have the algorithm for constructed.
The technology behind it is ultimately very simple (it is only math indeed), it can be done and it is done in other games.
In conclusion, a company like Blizzard that is very money-centric is not absurd that it is using a tool like this. I would not be surprised.
Pretty sure that it has! I'm getting tired of facing classes that perfect counters mine.
If the purpose of this algorithm is to force players to spend more money, then why has it been perfectly possible to hit legend with only face hunter since Descent of Dragons came out?
Well, this is because they had to balance between free-to-play and people who spend. Players are both. Anyway, it is not impossible to think that a face hunter player will find more matches against taunt druid, shadow priest or control warrior (that are not favourable matches) than a ramp druid for example.
These assumptions are pretty easy to check anyway. We need one person that plays around 100 matches or more with the same deck. Every deck is fine for this experiment.
If the numbers of his fights are aligned with the numbers in the attachment, everything is fine. If his numbers are not aligned with the numbers in the attachment we can assume that HS is controlled.
With an experiment like this, there is proof.
waiting for 3nnu1 to explain that the game whose forum he visits every day is rigged
You need two people at the same rank (and possibly same server - not entirely sure how much variance there is based on that) with roughly the same MMR, realistically - the large datasets show variance between ranks. In practice, this means two legend rank players with a similar rating and different decks.
Also, even just looking at purely the classes, I'd look at a bigger sample than 100 - more like 2-300. I presume you're not planning on looking at 40-50 deck types with 100 sample size, for what should be obvious reasons!
No. You need one person and the MMR is not relevant. You are testing the meta. Therefore if in 100 matches he will not find (with always the same deck) 23,6% of druid (or around this number) something is strange, otherwise, everything is fine.
But the meta will shift between different MMRs. Always has done. Top 100 legend meta will be significantly different to that in, say, silver 10 with a MMR that reflects that. Heck, it'll be different between bottom of legend more fun decks and D5-1 - in this case because of a difference in philosophy for the players.
People say that they play aggro until a certain point for faster and more consistent grinding, as an example.
The core issue is that you're matched based on MMR, which is invisible pre-legend - and that MMR influences your metagame, thus needs to be controlled for. 80%+ of actual science, in my experience, is controlling variables, cutting datasets, and similar.
You don't have to leave meta to shift. You have to try it during the same meta. If you let the meta shift during the experiment the whole experiment is pointless. Even though the experiment seems long, it is quite fast. Take a random deck (face hunter for example) and match the opponent and concede 100 times. Are your opponents aligned with 23,6% druids?
In Vicious syndicate, there are data about different ranks. So you have to choose one accordingly. Anyway, there are some differences, if you check the data, but nothing astonishing. 23,6% druid is the average in all ranks, so it is a reasonable benchmark.
https://www.vicioussyndicate.com/drr/classarchetype-distribution-data-reaper-report/
That theory is wrong as others have mentioned (it’s just you being on a losing streak). The way to confirm is: don’t switch decks. Run the same deck for 50-100 games and keep track of your matchups (use a deck tracker). It will prove to you that what you think is happening, is not actually happening.
HSReplay tracks every single game played by it's users. Well over a thousand games minute.
At no stage have they been able to track any kind of pattern in the match-ups and if there were any it would be quite obvious given how many games they record.
People continually post of this forum about being matched up badly when changing decks, seeing fixed RNG working against them, seeing card draws being either too lucky or unlucky to be believable...
But it's all based on annecdotes at worst, or a short run of tracked games in their personal log at best - neither of which amounts to dick all.
Of course activision Blizzard have IP on all sorts of algorithm, they are heavily in the loot box box business, across nearly all their games. That doesn't mean some of the more nefarious ones are in use in hearthstone - and the second they were people would see.... I mean, err, people that are not so susceptable to conspiracy and paranoia would see...the rest of you see it already when it isn't even there - there isn't much helping you, you've been told so many times now.
There are multiple difficulties with that approach.
Yes, you can't let the meta shift - but you are looking at 'all ranks' in the VS data. Click into the rank breakdown and you will see that the overall 23.6% is all over the place: from 15% at GSB to 38% at 1k legend. There is a 4% difference between the Diamond brackets.
So picking a random rank will not reproduce the same value.
You also cannot try to repeat last weeks results this week because the meta is shifting between weeks (which you can see in the different reports).
The meta is probably also shifting on a daily basis (weekdays - weekend), time of the day (ladder grinders during the day or casuals in the evening) and MMR (the skill to pilot certain decks).
So ideally you would want to break down the VS data to that level to have an accurate goal or in absence of that granularity, play enough games to cover all levels with a large enough sample size.
And then you have to consider the observer effect of you injecting games: if you beat players they might change their deck, if you concede they gain rank and move out of your bracket scope with possible knock-on effects. On top you are gaining or losing MMR, which can put you in a different meta.
So yeah, you want two players with different decks that queue at the same time into the same local meta to generate two data sets that you can then compare.
Yes. 4% difference is nothing. I don't know why you guys like complicated things. Mine is a pretty direct approach. No doubt that it is correct. Sure it is not perfect. But it is easy, correct and FAST (your models are not fast).
Sure if you make the model more complicated the result is better it is obvious but pointless. You can make it with 10 players and the model is even better and then?
Pretty easy and fast. Same deck. Concede 100 times. Are your data aligned with that of Vicious?
Blizz has a patent for these things, so this is an easy way to check if they are using it in HS constructed.
I don't know your experience in data analysis, I have it for big investment banks and I can tell you that 100 is a good sample size. Most of the decisions are generally taken with a lower sample size (because indeed most of the time there is no possibility to have a 100 sample size). So, nothing strange for me.
chill, man
you are so angry at everything that all your reasoning is biased and grounded on a prejudice ("Blizz... i hate them, they are the EVIL!!!"). you start with the conclusion and then try to find proofs to support your theory. scientific approach is the other way around mate. this, above, is called: paranoia.
"Woow..."
You can test it yourself relatively easily. Build a deck of absolutely bad cards, don't use any strategy. Just build a deck yourself. Your next opponent will also play an off-meta deck. Almost always works. It's just an extra screw to bring some "balance" into the game.
https://www.youtube.com/watch?v=2T658vTvoRs&t=581s&ab_channel=HearthstoneTipsbyKrisOFive
To live is to suffer, to survive is to find meaning in the suffer!
I am chill. We are talking about an approach to test it. You are out of topic. Please read better next time.
100 games (only concede) on silver 10: more Shamans and DH, almost no Rogues compared to VS's GSB ranking from last week. Took about an hour.
(I doubt that this will put the topic to rest though - because reasons.)
Or it could just be that people who are at a certain MMR... play off meta decks.
Thank you very much Banur for your effort.
Based on this data, it seems clear to me that the game is not rigged in the matchmaking system. It seems a closed point for me.